Prediction of cardiovascular disease risk among people with severe mental illness: A cohort study.

OBJECTIVE: To determine whether contemporary sex-specific cardiovascular disease (CVD) risk prediction equations underestimate CVD risk in people with severe mental illness from the cohort in which the equations were derived.
METHODS: We identified people with severe mental illness using information on prior specialist mental health treatment. This group were identified from the PREDICT study, a prospective cohort study of 495,388 primary care patients aged 30 to 74 years without prior CVD that was recently used to derive new CVD risk prediction equations. CVD risk was calculated in participants with and without severe mental illness using the new equations and the predicted CVD risk was compared with observed risk in the two participant groups using survival methods.
RESULTS: 28,734 people with a history of recent contact with specialist mental health services, including those without a diagnosis of a psychotic disorder, were identified in the PREDICT cohort. They had a higher observed rate of CVD events compared to those without such a history. The PREDICT equations underestimated the risk for this group, with a mean observed:predicted risk ratio of 1.29 in men and 1.64 in women. In contrast the PREDICT algorithm performed well for those without mental illness.
CONCLUSIONS: Clinicians using CVD risk assessment tools that do not include severe mental illness as a predictor could by underestimating CVD risk by about one-third in men and two-thirds in women in this patient group. All CVD risk prediction equations should be updated to include mental illness indicators.

Introduction

Experience of severe mental illness (SMI) is associated with higher prevalence, incidence and mortality from a range of cardiovascular diseases (CVD) including coronary heart disease, congestive heart failure and cerebrovascular disease.[1] This increased risk of CVD is an important factor in the high rates of premature mortality among people with SMI.[1,2] SMI can be defined narrowly to include people with diagnoses of functional psychosis including schizophrenia and bipolar disorder, or more broadly to also include major depression and/or anxiety, or using a definition which relates to the level of need for services or functional disturbance, with evidence of increased CVD risk in all groups.[1-4]

Established risk factors such as smoking, diabetes, and obesity partly explain the increased CVD risk.[5,6] However, there is evidence that the increased risk exceeds that due to established risk factors.[7,8] Possible explanations include biological factors related to mental illness, under-recognition of CVD leading to delayed diagnosis, and lack of appropriate primary and secondary preventative interventions.[9]

Risk prediction algorithms such as the Framingham Risk Score [10] are important for informing appropriate management of primary CVD risk. If mental illness is an independent risk factor for CVD then risk prediction algorithms based on established risk factors will underestimate risk. Current UK NICE guidance on cardiovascular risk assessment recognises but does not quantify this likely underestimation in people with additional risk due to antipsychotic medication or SMI.[11] The new QRISK3 score used in the UK includes SMI and atypical antipsychotic prescription as predictors to rectify underestimation of risk.[12] However, other CVD risk assessment algorithms currently being used (for example PREDICT in New Zealand[13]) do not include SMI, and so empirical investigation is needed to understand the magnitude of underestimation.[14]

The main aim of our study was to compare the observed risk of a first CVD event among people with SMI with the risk predicted by recently developed algorithms for the New Zealand general population,[13] in order to identify and quantify any underestimation. The hypothesis was that CVD risk prediction algorithms will underestimate the CVD risk for this group. SMI was defined as treatment by specialist mental health services in the five years prior to CVD risk assessment. A subgroup with diagnoses of schizophrenia or bipolar disorder were examined separately to investigate the utility of using wider vs narrower definitions of SMI.

Methods

Study population

This study uses the PREDICT cohort, a population-based anonymised cohort of people having their first cardiovascular risk assessment in primary care in New Zealand. The PREDICT cohort has been well described elsewhere.[13,15] Briefly, the cohort includes all people who have their CVD risk assessed in primary care using the PREDICT-CVD web-based tool, which is used by 35 to 40% of New Zealand general practices covering approximately 35% of the national resident population, mainly in the Northern part of New Zealand. The cohort is continually being updated. For the analyses presented here, CVD risk assessments between 20 October 2004 and 31 December 2016 were included, with data from 522,969 individuals. Routine CVD risk assessment was recommended from age 45 years (men) or 55 years (women) over the study period, and ten years earlier for people from Māori, Pacific and Indian subcontinent populations.[16] This cohort was linked to national pharmaceutical dispensing, hospitalisations, public mental health service records and mortality records using a unique national health identifier (the encrypted NHI). This study was limited to participants aged 30 to 74 years, which is the age group that the PREDICT CVD risk prediction algorithms were derived from. People with a history of prior CVD (including heart failure) or renal failure were excluded from the present study of primary risk prediction. Participants with missing data on predictor variables (2 with missing smoking status data values, 2659 with missing cholesterol data values) were excluded from the main analyses. Fig 1 details the cohort selection process.

Fig 1

Flow diagram for selection of cohort aged 30–74 without prior CVD from the PREDICT data set.

Maximum follow up time was 12.2 years, and the mean follow up time was 4.5 years. Because comprehensive routine national data on deaths and hospitalisations was used for follow up there was no loss to follow up, except for people who left the country.

PREDICT cohort variables

Exposure

People with SMI were the primary exposure group of interest in the current study and this patient group has not been previously identified within the PREDICT Cohort. Data on treatment by specialist mental health services were used to identify the population group with the most functionally disruptive mental illness. Treatment by specialist mental health services in the five years prior to the index CVD risk assessment was identified from face-to-face treatment contacts with mental health services recorded in the PRIMHD (Programme for Integration of Mental Health Data) dataset, which covers all public secondary mental health care in New Zealand.[2] Inpatient and face-to-face community based mental health treatment contacts in the five years prior to the index assessment date were identified from information on the type and setting of service activities.

Mental health diagnoses were identified from the PRIMHD dataset. Missing diagnosis is a substantial problem with the PRIMHD dataset, with 36.2% of people identified in our study as having treatment contact with mental health services between 2009 and 2016 but having no psychiatric diagnosis information recorded in PRIMHD. All available psychiatric diagnoses, including secondary and provisional, were included to maximise available information. For the analyses presented here those with diagnoses of functional psychosis (ICD codes F20-F31) were examined separately, as data on this diagnostic group are close to complete.[17] It was not possible to limit the overall definition of SMI to specific diagnoses because of missing diagnosis data, and so for this reason the whole group in contact with specialist services were defined as having SMI for the purposes of this study.

Predictors used to calculate CVD risk

Socioeconomic deprivation at the time of index assessment was measured using the New Zealand Deprivation Index (NZDep) 2013, which is an area-based measure of relative deprivation.[18]

We integrated multiple records of ethnicity from both the PREDICT dataset (recorded at primary care) and the National Health Index (recorded at secondary care) to ascertain a single prioritised ethnicity for each individual. The categories (in order of priority) are Māori (the indigenous population of New Zealand), Pacific, Indian, Chinese and other Asian, and residual group of other ethnicities (predominantly European).

The other predictors were standard CVD risk factors at baseline, which were drawn from data recorded by primary care clinicians at CVD risk assessment, augmented by prior hospitalisation, pharmaceutical dispensing, and lab test data. These risk factors were: age, gender, family history of premature cardiovascular disease, atrial fibrillation, diabetes, systolic blood pressure, TC:HDL ratio, and medications at index assessment (blood pressure lowering, lipid lowering, and antithrombotic medication). Definitions of these risk factor variables are available elsewhere.[16]

Outcomes

The primary outcome was total incident cardiovascular disease over the follow-up period, defined by ICD-10-AM codes as a death or hospitalisation from ischaemic heart disease, ischaemic or haemorrhagic cerebrovascular events, peripheral vascular disease, congestive heart failure or other ischaemic cardiovascular disease deaths.[13]

Analysis

Descriptive analyses of demographics and risk factors were performed, stratified by sex and SMI. Numbers and proportions are reported to enable comparison between groups. All those with SMI were compared to those without SMI. The subgroup with a history of schizophrenia or bipolar disorder diagnosis are also described separately. Risk factors were described for those aged 30 to 74 years at index risk assessment, stratified by SMI and presented as numbers and proportions.

Time-to-event curves, adjusted for age were used to compare the risk of CVD outcomes (fatal or non-fatal) between those with SMI and those without. Those who died from a non-CVD cause were censored at date of death. Among the group with SMI, those with diagnoses of schizophrenia or bipolar disorder and those without were also examined separately.

Observed and predicted risk among people with SMI were compared using calibration plots. Calibration plots are also presented for those without SMI. The PREDICT algorithms, which have been developed and validated for the New Zealand population,[13] were used to calculate the predicted risk in deciles. Kaplan Meier estimates of observed risk were derived from CVD events in the five years following index CVD risk assessment. Men and women were examined separately. The ratio of predicted to observed risk was calculated for each decile of risk and the mean ratio reported to enable quantification of any underestimation.

We used SAS 9.4 and R software for analyses.

Ethical approval

This study was approved by the New Zealand Northern A Health and Disability Ethics Committee, reference MEC/07/19/EXP/AM12. New Zealand ethics committees allow secondary re-use of health data without individual patient consent where data are not identifiable. Information about the PREDICT study is available at all general practice locations, and patients may opt out of having their de-identified data being included in the cohort.

Results

The PREDICT dataset was used to identify a cohort of 522,969 individuals who had a first CVD risk assessment between 2004 and 2016. This cohort was limited to 495,388 people aged 30 to 74 years at first risk assessment. Of these, 28,734 (5.8%) had also had face-to-face contact with specialist mental health services in the five years prior to their index CVD risk assessment, including 7669 (36.6%) women and 4,456 (15.5%) people with a recorded diagnosis of schizophrenia or bipolar disorder.

A total of 65,147 people were excluded from the analyses because of CVD or renal failure at the time of index risk assessment, leaving a final cohort of 430,241 individuals having a primary CVD risk assessment. Among women, 14.5% of those with SMI and 10.9% of those without SMI had prior CVD, while among men the proportions were 13.1% and 12.3% respectively.

Table 1 shows the demographic characteristics of the study participants aged 30 to 74 years without a prior history of CVD who had first CVD risk assessments over the study period. Those with SMI were younger at index assessment than those without SMI, and a higher proportion were Māori. SMI was associated with higher levels of deprivation, particularly among men and women with diagnoses of schizophrenia or bipolar disorder. In comparison to the total PRIMHD population, those who also appeared in the PREDICT dataset had a similar age and ethnicity distribution within the 45–74 year age group where routine screening was recommended (for example 49% aged 45–54 in PREDICT vs 53% in this age group in the total PRIMHD population aged 45–74; 26% Māori in PREDICT vs 24% in PRIMHD). There was no missing data on demographic variables.

Table 1

Demographic factors at baseline CVD risk assessment among people 30–74 years with no prior CVD, by prior mental health (MH) status and gender.

	Women						Men
	MH treatment past 5 years*		Schizophrenia/ Bipolar Disorder		No MH treatment past 5 years		MH treatment past 5 years*		Schizophrenia/ Bipolar Disorder		No MH treatment past 5 years
	n	%	n	%	n	%	n	%	n	%	n	%
Total	6544		1527		181272		11643		2341		230782
Age (years)
30–44	903	13.8	257	16.8	12683	7.0	4227	36.3	962	41.1	49912	21.6
45–54	2603	39.8	620	40.6	56167	31.0	5052	43.4	915	39.1	93885	40.7
55–64	2310	35.3	495	32.4	77881	43.0	1776	15.3	346	14.8	57649	25.0
65–74	728	11.1	155	10.2	34541	19.1	588	5.1	118	5.0	29336	12.7
Ethnicity
Maori	1805	27.6	475	31.1	23857	13.2	3674	31.6	831	35.5	25911	11.2
Pacific	586	9.0	196	12.8	22886	12.6	1679	14.4	341	14.6	27841	12.1
Indian	281	4.3	67	4.4	14463	8.0	498	4.3	76	3.2	20855	9.0
Other Asian	374	5.7	80	5.2	20581	11.4	397	3.4	86	3.7	23435	10.2
European/other	3498	53.5	709	46.4	99485	54.9	5395	46.3	1007	43.0	132739	57.5
Deprivation Quintile
1 (least deprived)	1045	16.0	141	9.2	40416	22.3	1294	11.1	153	6.5	52754	22.9
2	1053	16.1	186	12.2	35966	19.8	1541	13.2	250	10.7	46675	20.2
3	1169	17.9	239	15.7	32972	18.2	1959	16.8	404	17.3	41960	18.2
4	1332	20.4	360	23.6	33598	18.5	2526	21.7	524	22.4	41915	18.2
5 (most deprived)	1945	29.7	601	39.4	38320	21.1	4323	37.1	1010	43.1	47478	20.6

* the numbers in this column include all those using mental health services including those with a schizophrenia or Bipolar disorder diagnosis

Table 2 shows the distribution of cardiovascular risk factors and cardiovascular outcomes by SMI history for men and women. With the exception of smoking, there were no marked differences in risk factor distribution among those with SMI compared to those without. However, people who had a recorded diagnosis of schizophrenia or bipolar disorder had higher rates of diabetes, obesity and hypercholesterolaemia than those without a history of mental illness, and the differences were more marked among women.

Table 2

CVD risk factors at baseline CVD risk assessment and CVD events over follow up for people aged 30–74 years with no prior CVD, by prior mental health (MH) status and gender.

	Women						Men
	MH treatment past 5 years*		Schizophrenia/ Bipolar Disorder		No MH treatment past 5 years		MH treatment past 5 years*		Schizophrenia/ Bipolar Disorder		No MH treatment past 5 years
	n	%	n	%	n	%	n	%	n	%	n	%
Total	6544		1527		181272		11643		2341		230782
Family history of CVD	798	12.19	178	11.66	21433	11.82	1039	8.92	162	6.92	22682	9.83
History of diabetes	925	14.14	376	24.62	20952	11.56	978	8.40	401	17.13	21865	9.47
History of atrial fibrillation	52	0.79	10	0.65	1425	0.79	133	1.14	22	0.94	2777	1.20
Lipid lowering medication	920	14.06	300	19.65	28343	15.64	1295	11.12	423	18.07	34221	14.83
BP lowering medication	1471	22.48	364	23.84	47846	26.39	1626	13.97	365	15.59	44376	19.23
Antithrombotic medication	502	7.67	134	8.78	17031	9.40	675	5.80	162	6.92	20425	8.85
Past smokera	1061	16.21	203	13.29	26780	14.77	1947	16.72	303	12.94	42742	18.52
Current smoker	1903	29.08	554	36.28	20974	11.57	4837	41.54	1145	48.91	34877	15.11
Mean BMIb (kg/m2)	29.46	(SD 7.44)	31.67	(SD 7.86)	29.16	(SD 7.18)	29.51	(SD 6.29)	30.95	(SD 7.17)	29	(SD 5.64)
BMI <25	1613	24.65	256	16.76	45785	25.26	2135	18.34	362	15.46	42185	18.28
BMI 25–29	1547	23.64	342	22.40	44508	24.55	3417	29.35	623	26.61	78232	33.90
BMI 30–34	1078	16.47	301	19.71	27651	15.25	2345	20.14	527	22.51	42332	18.34
BMI 35–39	599	9.15	199	13.03	14922	8.23	900	7.73	236	10.08	15438	6.69
BMI 40+	476	7.27	191	12.51	11815	6.52	571	4.90	204	8.71	8116	3.52
Missing	1231	18.81	238	15.59	36591	20.19	2275	19.54	389	16.62	44479	19.27
Mean TC:HDLc	3.86	(SD 1.23)	4.14	(SD 1.37)	3.71	(SD 1.08)	4.45	(SD 1.42)	4.79	(SD 1.56)	4.39	(SD 1.24)
TC:HDL >4	2389	36.51	689	45.12	58941	32.52	6670	57.29	1524	65.10	131330	56.91
Mean SBP, mmHg	125.65	(SD 17.55)	123.92	(SD 16.79)	128.62	(SD 17.59)	127.1	(SD 16.27)	124.36	(SD 15.81)	128.82	(SD 16.11)
Mean DBP, mmHg	78.06	(SD 10.76)	77.7	(SD 10.46)	78.4	(SD 10.15)	80.09	(SD 10.89)	78.97	(SD 10.42)	80.07	(SD 10.23)
Elevated BPd	3528	53.91	756	49.51	109880	60.62	6739	57.88	1211	51.73	143798	62.31
CVD events over follow up	241	3.68	68	4.45	5880	3.24	448	3.85	88	3.76	9883	4.28

aSmoking status missing data values on 2 patients

bBMI missing data values for 84,576 patients (19.7%)

cTC:HDL ratio missing values for 3388 patients (0.8%)

SBP>120 mmHg or DBP>90 mmHg

* the numbers in this column include all those using mental health services including those with a schizophrenia or Bipolar disorder diagnosis

Fig 2 shows age-adjusted estimates of the risk of a CVD event over the first 8 years of follow up time stratified by history of mental illness. Among both men and women, a history of SMI is associated with an increased risk of a cardiovascular event. This was the case for people with diagnoses of schizophrenia or bipolar disorder and also for others using specialist mental health services, with overlapping confidence intervals for the two subgroups (see Fig 3).

Fig 2

Age adjusted time to event plots of risk of cardiovascular event by prior mental illness (SMI), limited to people aged 30–74 years at index assessment with no prior CVD, dashed lines indicate 95% confidence limits (n = 430241, events = 17197).

Fig 3

Age adjusted time to event plots of risk of cardiovascular event across three categories of prior mental illness (SMI), limited to people aged 30–74 years at index assessment with no prior CVD, dashed lines indicate 95% confidence limits (n = 430241, events = 17197).

Figs 4 and 5 compare observed 5-year risk (x axis) with deciles of predicted 5-year risk (y axis) of CVD events in the total study population aged 30 to 74 years. Estimates sitting on the diagonal line indicate no difference between observed and predicted risk per decile of the cohort (i.e. accurate risk prediction by the algorithm). Fig 4 compares people with SMI (left) and those without a history of mental health service use (right). For people with SMI, observed risk is higher than predicted risk across deciles of risk, indicating that the risk prediction algorithm is underpredicting the risk of CVD events. For example, in the highest decile of risk, the predicted risk of a CVD event was approximately 11% over 5 years, while the observed risk was approximately 14%. The same pattern was found in both men and women, although more pronounced among women (see Fig 5). The mean ratio of observed to predicted risk is 1.64 for women and 1.29 for men, and for men and women combined is 1.37. For those without a history of mental health service use the observed and predicted risks are approximately equal.

Fig 4

Predicted vs observed CVD risk among people aged 30–74 years with no prior CVD, with SMI (left), and without SMI (right) Error bars indicate 95% confidence limits. Blue diagonal line indicates observed = predicted risk (n = 426911, events = 16289).

Fig 5

Predicted vs observed CVD risk among men (left) and women (right) with history of prior mental illness (SMI) in the five years prior to index assessment. Error bars indicate 95% confidence limits. Blue diagonal line indicates observed = predicted risk (n = 18055, events = 680).

Discussion

We found that among people aged 30 to 74 years without a history of CVD, who had a cardiovascular risk assessment in primary care, those with a history of SMI tended to be younger, more likely to be Māori, and live in more deprived areas. They also had higher smoking rates, although other risk factors were similar, except among the subgroup with schizophrenia or bipolar disorder who had higher rates of metabolic disturbances. The age-adjusted risk of CVD events was elevated in those with SMI, among those both with and without diagnoses of schizophrenia or bipolar disorder. When CVD risk predicted by the contemporary PREDICT algorithm was compared to the observed risk over five years, the algorithm consistently underestimated observed risk among both men and women with SMI, particularly in the top five deciles of predicted risk.

These findings demonstrate that risk factor patterns including metabolic disturbances may only be part of the reason for the high rates of CVD among people with SMI and point to other factors such as overshadowing of physical health problems by mental health problems leading to delayed diagnosis and differences in the quality of preventive care received by those with mental illness compared to those without [4,9]. These mechanisms are likely to operate more strongly for more severe and stigmatised diagnoses and may partly explain the differences between those with and without functional psychosis.

Strengths and limitations

We used a large primary care database with near complete coverage of the eligible population in the region in general practices using the PREDICT software, linked to data on specialist mental health service treatment to identify a history of SMI. Public specialist mental health services in New Zealand provide inpatient and community treatment for the approximately 3.5% of the population with the highest mental health need, making this an appropriate method for identifying a cohort with the most functionally disruptive mental illness.[19] People with diagnoses of functional psychosis or severe depression who have used specialist services more than five years previously or are being cared for by private services or primary care will not be identified as having SMI in this study. However, there is little private treatment for severe mental illness in New Zealand, and most people unwell with severe disorders use specialist care regularly, meaning that coverage of those with the most severe disorders will be reasonably complete. Those who are missed are likely to have less severe or disruptive conditions and so this method may overestimate the difference in cardiovascular risk between those with and without SMI if more severe illness is associated with a higher risk of CVD (as is suggested by the higher risk among those with psychosis diagnoses in this study). On the other hand, this method of identifying SMI will also include people with diagnoses not generally included in definitions of SMI, as well as people who have a relatively short duration of mental illness, and these groups may have a lower CVD risk than those with severe depression and psychosis, leading to a potential underestimate of the difference between those with and without SMI. Hence our method of ascertainment is imperfect but is likely to identify an unbiased population of those with SMI. Moreover, the separate analysis of those with psychosis provides a more precise and reproducible narrow definition of SMI. New Zealand has a similar patterns of both cardiovacular disease and mental illness to other high income countries and so these results are likely to be generalisable.

We only included people who had CVD risk assessed in primary care, and therefore run the risk of recruitment bias. Overall, 90% of eligible patients in the Auckland and Northland regions are included in PREDICT,[15] but it is not known whether a similar proportion of those with SMI are included. International data suggests lower rates of CVD risk assessment among those with SMI than the general population.[20] Moreover, a lack of clear demarcation of roles mean that some of those accessing secondary mental health services will be having their CVD risk assessed and managed by mental health services. Nevertheless, those who are not included in the study because they are missing out on CVD risk assessment or accessing CVD risk assessment elsewhere are likely to be at higher risk than those captured in routine primary care risk assessments, and so any bias would be in the direction of underestimating the difference in risk between those with and without mental illness.

One of the major strengths of the PREDICT cohort is the near complete availability of cardiovascular risk factor and demographic information for the cohort from primary care data, and of cardiovascular events through linkage to national hospitalisation and mortality data. Completeness of demographic and risk factor data was equal between those with and without prior mental illness. It is possible that there is differential under-ascertainment of cardiovascular events among those with severe mental illness due to diagnostic overshadowing,[21] which would result in an underestimate of the risk of non-fatal cardiovascular events. Again, this bias would result in an underestimation of the difference between those with and without mental illness.

It is likely that given the extent of missing psychiatric diagnosis data, a small proportion of those without any recorded diagnosis did in fact have a diagnosis of functional psychosis. However, this was minimised by the inclusion of all provisional and primary psychosis diagnoses, including those recorded after the index assessment. Any misclassification would result in underestimating the difference between the two subgroups with SMI, but would not affect the main analyses which combined these groups.

Strengths and weaknesses in relation to other studies

Other studies have compared predicted CVD risk between those with and without SMI, with many finding little difference in predicted risk between groups.[22-24] Although these studies point to underestimation, they were not able to confirm this as they did not include outcomes data. In contrast, our study included individual level outcome data so it was possible to investigate the performance of the algorithm to confirm and quantify underestimation. Two previous studies have also used both risk factor and outcome data for this population to understand increased risk and are discussed below.

Osborn and colleagues directly compared observed and predicted risk and found that the Framingham algorithm overestimated the risk of CVD events for people with SMI (as it does for the general population), and that Framingham recalibrated for the UK population underpredicted risk in women but not men with SMI.[25] This is consistent with our finding of a higher ratio of observed to predicted risk among women with SMI than men. However, this study did not include a comparison population without SMI, and did not use an up to date algorithm which would be recommend in practice.

In developing the UK based QRISK3 risk assessment algorithm, investigators found that both SMI (defined as schizophrenia or bipolar disorder) and atypical antipsychotic prescriptions independently predicted CVD outcomes, with an adjusted hazard ratio of 1.13 (men and women) for SMI and 1.29 (women) and 1.14 (men) for antipsychotics. Taken together, the magnitude of increased risk attributed to these two factors is comparable with the ratio of observed to predicted risk found in our study, despite the different definition of SMI used. Of particular note is the greater degree of increased risk among women with a history of mental illness compared to men with a similar history, which again was consistent with our findings.

A first cardiovascular risk assessment appeared to be done at a younger age among those with a history of SMI, although the pattern seen is also likely to relate to the young age distribution of those in contact with mental health services. Within the 45 to 74 years age range there was no evidence of earlier risk assessment among those with SMI, with the age distribution of people in PRIMHD and PREDICT mirroring the age distribution in PRIMHD. The recently updated New Zealand CVD consensus statement [26] recommends that people with SMI (defined as schizophrenia, bipolar affective disorder or major depression) have their CVD risk assessed regularly from the age of 25. This is an earlier age than those included in this study, and further work is needed to assess the impact and predictive power of risk assessment from this age.

We have confirmed that even the most up to date and well calibrated CVD risk algorithm for primary prevention substantially underestimates the risk for those with SMI. Importantly, this underestimation is not limited to those with a diagnosis of functional psychosis, who have been the focus of previous studies.[25] It is therefore appropriate that those with major depression should be included in the definition of SMI for the purposes of identifying those at increased risk of CVD, as proposed in the NZ CVD consensus statement.[26]

Conclusions

Implications for practice

Demonstrating the magnitude of this underestimation of CVD risk is important for primary care practice, as mental illness is not specifically included most available risk prediction algorithms. The observed risk of an event in five years was 60% higher than estimated by the algorithm for women and 30% higher than estimated for men. Therefore, primary care providers using most common CVD risk prediction algorithms to inform management decisions will need to adjust the calculated risk upwards.

Unanswered questions and future research

This study provides a clear rationale for the development of CVD risk prediction algorithms that include predictors for people with SMI. The updated PREDICT algorithm is in the process of being made available for practitioners in New Zealand and this presents an opportunity to include either a separate model for SMI or SMI as an additional predictor. Further investigation is needed to understand the reasons for the higher risk of CVD over and above that predicted by established risk factors. In particular, modifiable factors such as diagnostic overshadowing and differential receipt of treatment, need to be investigated to inform interventions to improve outcomes.

1 Jul 2019

PONE-D-19-15070

Prediction of cardiovascular disease risk among people with severe mental illness: a cohort study

PLOS ONE

Dear Dr. Cunningham,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The two reviewers addressed a number of major and minor concerns about your manuscript. Please revise your manuscript carefully.

We would appreciate receiving your revised manuscript by Aug 15 2019 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Kenji Hashimoto, PhD

Academic Editor

PLOS ONE

Journal Requirements:

1. When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Thank you for stating that your study was approved by the New Zealand Northern A Health and Disability Ethics Committee, reference MEC/07/19/EXP/AM12.

Please place this information in your methods section.

3. Thank you for stating that the dataset was anonymised. Please provide this information in your ethics statement in the manuscript and in the online submission form. Please also clarify whether the ethics committee waived the requirement for informed consent. If patients provided informed written consent to have data from their medical records used in research, please include this information.

4. We note that you have included the phrase “data not shown” in your manuscript. Unfortunately, this does not meet our data sharing requirements. PLOS does not permit references to inaccessible data. We require that authors provide all relevant data within the paper, Supporting Information files, or in an acceptable, public repository. Please add a citation to support this phrase or upload the data that corresponds with these findings to a stable repository (such as Figshare or Dryad) and provide and URLs, DOIs, or accession numbers that may be used to access these data. Or, if the data are not a core part of the research being presented in your study, we ask that you remove the phrase that refers to these data.

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This is a well-written paper and an important topic. The authors leveraged unique data sources to investigate CVD risk in those with severe mental illnesses (SMI). There are some concerns of the manuscript that limit its impact in its current form, as discussed in detail below. The three major concerns are the lack of clarity on the study population, the definition of SMI, and the lack of a proposed mechanism of the association between SMI and CVD events. The total number of participants in the PREDICT study was not provided nor was the way the investigators ended up with the analytic study population. It would help to show a diagram of the data sources and how the participants flowed into the final study population. The methods on page 4, line 84 state that data from 522,969 participants were included, but the tables show ~434,109. The other concern is with the ascertainment and definition of SMI. Could the authors comment on how accurate contact with specialist mental health services correlate to a SMI? It could be possible that people have depression or another SMI, but do not seek help from a specialist. Additionally, the definition of SMI is general and the diagnoses are missing. Both of these concerns could lead to misclassification of the exposure. If the authors conclude that SMI should be incorporated in CVD risk prediction tools, what variable should the tools use given the lack of SMI definitions in this analysis. Lastly, the paper would benefit from a proposed mechanism, which could vary depending on the specific SMI.

Minor areas of improvement for the authors to consider that could strengthen the paper are discussed below.

1. Do the authors compare the observed risk using the QRISK3 score that includes SMI?

2. Was there perfect linkage between the data sources?

3. The authors mention there was no loss to follow up, except for those who left the country. Could the authors provide the N?

4. How many events occurred and what are the sample sizes for the estimates in the figures?

5. Figure 1 shows the patterns of CVD events over 8 years. What is the rationale for showing 8 years?

6. Including those who had contact with a specialist mental health service, but did not have a diagnosis of a psychotic disorder is confusing. As mentioned, it is due to lack of data, but labeling those individuals as having a SMI is problematic.

Reviewer #2: This is a valuable article investigating the prediction of cardiovascular disease risk among people with severe mental illness. The paper reports that the PREDICT equations underestimated the risk for this group, with a mean observed: predicted risk ratio of 1.29 in men and 1.64 in women. This is a valuable paper that presents evidence of underestimating risk of CVD risk assessment tools in patients with severe mental illness.

I have the following concerns.

1. Discussion. P15, line 263. “hazard ratio of 1.[13]”

I think square brackets are unnecessary.

2. Discussion.

Whereas updated QRISK3 considers atypical antipsychotics and severe mental illness as risk factors, QRISK2 did not include them. I want to know if New Zealand people have plans to upgrade the PREDIC algorithm in the near future. If so, the findings of this study will be a great help.

In conclusion, I enjoyed reading this paper. This is a valuable paper that presents evidence of underestimating risk of CVD risk assessment tools in patients with severe mental illness.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

5 Aug 2019

Please note this response is also attached as a separate document

Response to reviewers:

Prediction of cardiovascular disease risk among people with severe mental illness: a cohort study

Thank you for the opportunity to submit a revised version of our manuscript and respond to the reviewers comments.

We have revised the manuscript and attach a tracked changes and clean copy of the revised manuscript.

We have included the requested extra information on the ethical approval process in the methods section of the paper and performed the requested formatting changes.

Below we have responded specifically to the comments from the reviewers (responses in red). Note: the line numbers refer to the tracked changes version of the revised paper.

Reviewer #1: This is a well-written paper and an important topic. The authors leveraged unique data sources to investigate CVD risk in those with severe mental illnesses (SMI). There are some concerns of the manuscript that limit its impact in its current form, as discussed in detail below. The three major concerns are the lack of clarity on the study population, the definition of SMI, and the lack of a proposed mechanism of the association between SMI and CVD events.

The total number of participants in the PREDICT study was not provided nor was the way the investigators ended up with the analytic study population. It would help to show a diagram of the data sources and how the participants flowed into the final study population. The methods on page 4, line 84 state that data from 522,969 participants were included, but the tables show ~434,109.

Response:

We have amended the first paragraphs of the methods and results sections to make it clear that the initial sample (n=522,969) was restricted by age (n=495,388) and then further restricted by removing those with prior CVD to produce the final analytic sample (n=430,241). Note that the group with schizophrenia and bipolar disorder are a subset of the group who have had prior contact with mental health services, and so the total shown in table 1 and 2 is 430,241 not 434,109. We have added footnotes to the tables to clarify this. A flow diagram is also provided (new figure 1).

The other concern is with the ascertainment and definition of SMI. Could the authors comment on how accurate contact with specialist mental health services correlate to a SMI? It could be possible that people have depression or another SMI, but do not seek help from a specialist. Additionally, the definition of SMI is general and the diagnoses are missing. Both of these concerns could lead to misclassification of the exposure. If the authors conclude that SMI should be incorporated in CVD risk prediction tools, what variable should the tools use given the lack of SMI definitions in this analysis.

Response:

Thank you for this comment. Our aim in this paper was to estimate the impact of mental illness on cardiovascular risk not only among those with a psychosis diagnosis but also among others with severe and persistent mental illness. This is important because of the evidence that non-psychotic disorders such as severe depression are also associated with an increased risk of CVD (for example Correll 2017).

We have used contact with specialist mental health services as an indicator of SMI. The publically funded secondary mental health service in New Zealand aim to provide care for the 3% of the population with the highest mental health needs. The threshold for entry is mental illness that is too significant and too complex to be managed in primary care such that it requires specialist input. Therefore almost all of those accessing secondary mental health services in New Zealand will have serious mental illness.

It is worth noting that unlike most other Western countries, New Zealand does not have a well-developed private psychiatry workforce. For example, there is only one small private (not for profit) inpatient facility in the country, but even in this facility 2/3 of the beds are referred and funded through the secondary mental health services.

However, as the reviewer correctly comments, using contact with secondary mental health services as the exposure will not detect all people with SMI, as those with depression or another SMI may not seek help at all, or may be managed in primary care. It will also include people who would not be captured by a different definition of SMI which relied on diagnosis or duration of illness.

This is a limitation of our study. The study population represents a large cohort of people with SMI, but will not include everyone in the country with SMI. We have included further comment on this limitation in our methods (line 117–9) and the discussion (line 253–61), as well as comment on what this means for incorporating markers on severe mental illness in any risk prediction tool (line 327-9).

Lastly, the paper would benefit from a proposed mechanism, which could vary depending on the specific SMI.

Response:

As discussed in the second paragraph of the introduction to the paper, a number of possible explanations have been proposed for the increased risk of CVD among those with severe mental illness over and above that due to established risk factors including biological factors related to mental illness, under-recognition of CVD leading to delayed diagnosis, and lack of appropriate primary and secondary preventative interventions. We have included comment about potential mechanisms in the discussion (line 236).

Minor areas of improvement for the authors to consider that could strengthen the paper are discussed below.

1. Do the authors compare the observed risk using the QRISK3 score that includes SMI?

Response:

We are unable to compare the observed risk with the QRISK3 score in this paper. The QRISK3 algorithm includes a number of factors which we do have recorded for our cohort. These include history of migraines, systemic lupus erythematosus, and erectile dysfunction.

2. Was there perfect linkage between the data sources?

Response:

The data are linked using the unique identifier (NHI) which is present on all health service records, and which over 98% of New Zealanders have. It is increasingly rare for an individual to have more than one NHI however when duplicate records are identified (in the course of healthcare provision or during Ministry of Health audit), one of their NHI numbers is deemed the primary and others become linked to that as secondary NHI numbers. Linkage across data sources therefore recognises where individuals may have more than one NHI, meaning the risk of imperfect linkage is extremely minimal.

3. The authors mention there was no loss to follow up, except for those who left the country. Could the authors provide the N?

Response:

Emigration status is not recorded in national health records so we are unable to estimate how many in the cohort may have left the country during the follow-up period of these data. We know that almost 34,000 New Zealand citizens left New Zealand in the year ending July 2018. If emigration is evenly distributed nationally it would be expected that 1/3rd of this number resided in the area covered by this study. We do not know how many of these were in our age range of 30-74 years, nor how many are likely to have severe mental illness.

4. How many events occurred and what are the sample sizes for the estimates in the figures?

Response:

We have included the event numbers and total number of observations for each figure in the figure captions, and in the figures for the survival plots.

5. Figure 1 shows the patterns of CVD events over 8 years. What is the rationale for showing 8 years?

Response:

Eight years represented the 75th centile of the distribution of duration. The mean was 4.5 and maximum was 12.2 years, and so the inclusion of the final four years of data did not add extra information as it represented few people, and yet limiting to 5 years, for example, would misrepresent the usable information we have.

6. Including those who had contact with a specialist mental health service, but did not have a diagnosis of a psychotic disorder is confusing. As mentioned, it is due to lack of data, but labelling those individuals as having a SMI is problematic.

Response:

As noted above, the reason for including those without a diagnosis of a psychotic disorder is in order to investigate the relationship between mental illness and cardiovascular risk prediction accuracy for the broader group with mental illness at increased risk of CVD. We have clarified this and the limitations of our approach in the discussion of the paper.

Reviewer #2: This is a valuable article investigating the prediction of cardiovascular disease risk among people with severe mental illness. The paper reports that the PREDICT equations underestimated the risk for this group, with a mean observed: predicted risk ratio of 1.29 in men and 1.64 in women. This is a valuable paper that presents evidence of underestimating risk of CVD risk assessment tools in patients with severe mental illness.

I have the following concerns.

1. Discussion. P15, line 263. “hazard ratio of 1.[13]”

I think square brackets are unnecessary.

Response:

This was an error and we have made this change.

2. Discussion.

Whereas updated QRISK3 considers atypical antipsychotics and severe mental illness as risk factors, QRISK2 did not include them. I want to know if New Zealand people have plans to upgrade the PREDIC algorithm in the near future. If so, the findings of this study will be a great help.

Response:

The PREDICT algorithm used here is in the process of being made available for clinicians to use. There are plans to provide an adapted PREDICT algorithm which takes account of the increased risk among people with SMI. Our group are in discussions with the providers of the PREDICT software about how best to do this. We have included a sentence about this in the paper conclusions (line 340).

In conclusion, I enjoyed reading this paper. This is a valuable paper that presents evidence of underestimating risk of CVD risk assessment tools in patients with severe mental illness.

Thank you for the opportunity to submit a revised version of this paper for consideration for publication.

Yours sincerely,

Ruth Cunningham

Submitted filename: response to reviewers PlosOne final.docx

Click here for additional data file.

9 Aug 2019

Prediction of cardiovascular disease risk among people with severe mental illness: a cohort study

PONE-D-19-15070R1

Dear Dr. Cunningham,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

With kind regards,

Kenji Hashimoto, PhD

Section Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors were responsive to the reviewer comments. Since there was excellent data linkage, one suggestion would be to elaborate on it more as a strength on page 16, line 269.

Reviewer #2: I believe the paper will be of interest to the readership of Plos One and would recommend it for acceptance.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

27 Aug 2019

PONE-D-19-15070R1

Prediction of cardiovascular disease risk among people with severe mental illness: a cohort study

Dear Dr. Cunningham:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Prof. Kenji Hashimoto

Section Editor

PLOS ONE